3406

Jpn. J. Appl. Phys. Vol. 39 (2000) pp. 3406–3409Part 1, No. 6A, June 2000c©2000 The Japan Society of Applied Physics

Low Threshold GaInAsP Lasers with Semiconductor/Air Distributed Bragg ReflectorFabricated by Inductively Coupled Plasma EtchingMaiko ARIGA, Yushi SEKIDO, Atsushi SAKAI , Toshihiko BABA, Akihiro M ATSUTANI1,Fumio KOYAMA 1 and Kenichi IGA1

Yokohama National University, Division of Electrical and Computer Engineering, 79-5 Tokiwadai, Hodogayaku, Yokohama, 240-8501, Japan1 Tokyo Institute of Technology, Precision and Intelligence Laboratory, 4259 Nagatsuda, Midoriku, Yokohama, 226-8503, Japan

(Received October 22, 1999; accepted for publication March 14, 2000)

We fabricated GaInAsP/InP short cavity lasers with semiconductor/air distributed Bragg reflectors (DBRs) by inductivelycoupled plasma etching with pure Cl2 gas. Nearly vertical sidewalls with low roughness of∼10 nm were achieved, separated byair spaces of three quarter wavelengths. The lowest threshold current normalized by the stripe width was 3.2 mA/µm. From thisvalue, the DBR reflectivity was evaluated to be 85%, which agreed with the theoretical value obtained from a finite-differencetime domain (FDTD) simulation. We compared two types of devices with different DBR shapes, and observed that DBRreflectivity was affected more by the tilt of the DBR sidewalls than the sidewall roughness. This result also agreed well withthe FDTD theory.

KEYWORDS: semiconductor laser, GaInAsP/InP, distributed Bragg reflector laser, inductively coupled plasma etching, finitedifference time domain simulation

1. Introduction

Laser diodes for access networks and local area data linksrequire simple fabrication processes for easy mass productionand high performance, as well as the ease of integration withother elements, for example, waveguides, detectors and fil-ters. A short cavity laser with a semiconductor/air distributedBragg reflector (DBR)1–6) is suitable for this purpose. It con-sists of a stripe cavity with a GaInAsP strained quantum wellactive layer and semiconductor/air DBRs, which are semicon-ductor vertical walls separated by air spaces. Such a structurecan be monolithically formed by anisotropic etching. Thelarge refractive index contrast allows a high reflectivity evenwith a small number of periods, so that a short cavity is possi-ble, which gives a low threshold current suitable for zero-biasmodulation and single longitudinal mode operation.

However, the fabrication of such DBRs, which are made ofInP-based materials, is a challenge. So far, dry and wet etch-ing processes have been used. Using Cl2 reactive ion beametching (RIBE), vertical sidewalls were easily achieved, butthey were rough and nonuniform.1,2) Wet etching realizedsmooth sidewalls, but they were severely tilted.4) In addition,the mechanism of reflectivity degradation was not clear. Inthis study, we used inductively coupled plasma (ICP) etchingto solve this problem. Using this method, fine and smoothetching has been demonstrated in this material system.7) Wefabricated the DBR and evaluated its reflectivity based on itslasing characteristics. We compared the evaluated reflectiv-ity with the calculated reflectivities obtained using the two-dimensional finite difference time domain (FDTD) method,and investigated the dominant factor required for a high DBRreflectivity.

2. Fabrication of DBR

We prepared an epitaxial wafer grown on n-InP substrateby metal organic vapor phase epitaxy. Epilayers consistedof an n-InP cladding layer, 0.25-µm-thick GaInAsP activelayer, 1.7-µm p-InP cladding layer and 0.3-µm p-GaInAscontact layer from the bottom. The active layer consistedof eight compressively-strained quantum-wells of∼4 nmthickness each with 10-nm-thick strain-compensated barriers

During the fabrication process, AuZn and AuGe electrodeswere formed on the epitaxial and back surfaces of the wafer,respectively. After patterning the AuZn electrode into a rect-angular shape, the DBR pattern was drawn on this side byelectron beam lithography using negative resist SAL601-SR7(Shipley Co., Inc.), and the epilayers were etched using theICP etching machine RIE-101ip (SAMCO International Inc.)with pure Cl2 gas. The gas flow rate was 10.0 sccm and thegas pressure was 0.5 Pa. The ICP power and the bias powerwere 300 W and 200 W, respectively. Under these conditions,the etch rate of InP was nearly 1.0µm/min. Figure 1 showsthe scanning electron micrograph (SEM) of the sample. Thesidewall angle was more than 89◦ and the sidewall roughnessestimated roughly was less than 10 nm.

3. Lasing Characteristics

For measurement purposes, the wafers were cut into pieces,as each of them had one device chip. Each piece was placedn-side down on a submount without a metal bonding. Fig-ure 2 shows typical lasing characteristics of a sample of cav-

and gradient-index separate confinement heterostructure lay-ers. The peak wavelength of the spontaneous emission was1.53µm, and the lasing wavelength of the broad area laserswas typically 1.55µm. The design and the fabrication pro-cess were similar to that reported previously1,2) except for theetching method. The width of the stripew was 20µm. Thecavity lengthL, i.e., length of the stripe, was varied from 150to 250µm in the samples. The DBR was placed at one end ofthe stripe as a backside mirror, and for the other end, a simpleetched facet was used as a front mirror. The width of the DBRwas designed to be twice the stripe width, i.e., 40µm, to avoidthe influence of fragile side edges of the DBR. The thick-nesses of the semiconductor walls and air spaces were threequarters the lasing wavelengthλ in the media, i.e., 0.35µmand 1.15µm, respectively, forλ = 1.55µm. The number ofsemiconductor/air pairsN was fixed at 4. According to theFDTD calculation of the DBR reflectivity,N = 3 is sufficientto obtain the maximum reflectivity limited by the diffractionloss in air spaces. However, a largerN was effective for re-ducing the nonuniformity of the thickness of the semiconduc-tor walls.

ity lengthL = 250µm at room temperature under the pulsedcondition (pulse width of 200 ns and repetition frequency of2 kHz). In Fig. 2, the current is normalized by the stripewidthw. The lowest normalized threshold currentI th/w was3.2 mA/µm. This value is nearly the same as that recordedfor anL = 100µm device fabricated by RIBE etching.2) Thespectra show multimode lasing characteristics, as shown inthe inset of Fig. 2. This is due to the wide stopband of thistype of DBR. The single mode operation can be obtained byreducing the cavity length to less than 50µm.

We observed a clear relationship between the DBR shapesand the etched facet at the other end, and the lasing charac-teristics. Using the different chamber conditions, we obtainedtwo types of devices with different shapes of DBR. The las-ing characteristics of these devices and SEM photographs aresummarized in Fig. 3. Device A had an nearly vertical semi-conductor sidewalls with relatively high roughness of over20 nm. Device B had tapered sidewalls with low roughnessof ∼10 nm. The shapes and roughnesses of the etched facetsin these devices were almost the same as those of the DBR. It

is clear that the threshold current was much lower for deviceA than device B.

In Fig. 4, we plot the measured threshold current densityJth with the inverse cavity lengthL−1. Straight lines indicatetheoretical values obtained by assuming the logarithmic gaing = g0 ln(J/J0) for strained quantum-wells, and the currentdensity for transparencyJ0 = 512 A/cm2, a gain coefficientg0 = 627 cm−1 and an internal cavity loss of 5 cm−1, whichwere evaluated for cleaved lasers. Experimental data for cav-ity lengths show relatively large dispersion values. This ismainly due to the irregular scattering loss at stripe mesas,etched facets and DBRs, which are mechanically damageddue to the cleaving of the wafer and mounting of each devicechip on a submount with a metal fitting. We estimated the

Current Normalized by Stripe Width [mA/µm]

50 10 15 20

Ligh

t Out

put [

a.u.

]

RT, Pulse(200ns, 2kHz) Wavelength [µm]

1.555 1.58 1.605

2.52

1.70

1.441.28

I /Ith=

Fig. 2. Light output versus normalized current characteristics and lasingspectra measured for a device with cavity length of 250µm.

Fig. 1. Side view of fabricated DBR.

Ligh

tOut

put[

arb.

units

]

Current Normalized Stripe Width [mA/µm]100 20 30 40 50 60

Device A

Device B

1µm

1µm

Fig. 3. Side views and lasing characteristics of two types of devices Aand B.

20

706050403020100

150200250400

10

2

5

1

0.5

Cavity Length L [µm]

Inverse Cavity Length L−1 [cm−1]

Thr

esho

ld C

urre

nt D

ensi

ty

J th

[kA

/cm

2 ]

Fig. 4. Threshold current density versus inverse cavity length characteris-tics. Open and closed circles indicate devices A and B, respectively. Solidline indicates theoretical results obtained by assuming DBR reflectivityRDBR of 85% and facet reflectivityRf of 28%. The dashed lines indicateresults whenRDBR = 50% andRf = 10% are assumed.

Jpn. J. Appl. Phys. Vol. 39 (2000) Pt. 1, No. 6A M. ARIGA et al. 3407

giving the peak reflectivity, as shown in Fig. 7. The reflectiv-ity is strongly affected by the tilt of the sidewall; a tilt of only5◦ reduces the reflectivity to∼30%.

We also investigated the behavior of light around the DBRin device A by the FDTD method. The light pulse reflectedby the DBR is displayed in Fig. 8. From this figure, we es-

3408 Jpn. J. Appl. Phys. Vol. 39 (2000) Pt. 1, No. 6A M. ARIGA et al.

evaluated to be 85%. On the other hand, the reflectivity was50% for device B. Differences in the facet and DBR reflec-tivities indicate that the reflectivity is strongly affected by thetilt of the semiconductor sidewalls and is not very sensitive tothe sidewall roughness.

4. Considerations

We compared the obtained reflectivities with those calcu-lated by the FDTD method.8) For the calculation, we usedYee’s 20-nm-square cells, dividing the model of the laserstructure viewed from the lateral direction, as shown in Fig. 5.We assumed an infinite lateral width and a transverse electricfield of the laser mode. A Gaussian pulse of 40 fs full widthat 1/e2 maximum with the guided mode profile in the laserwaveguide was used as an excitation field. The DBR reflectiv-ity was evaluated by calculating the overlap integral betweenthe profile of the analytical guided mode in the waveguide andthat of the light pulse reflected by the DBR and coupled againto the waveguide. We input the measured data of the DBRshapes of devices A and B into a computer. Figure 6 showsthe reflection spectra of the DBRs, which were obtained bythe Fourier transform of the time series of the overlap inte-gral. For these devices, the stopband lies within 1.3µm to1.7µm. The peak reflectivity is at a shorter wavelength thanthe stopband center, since the longer wavelength light suffersa larger diffraction loss in air spaces. Experimental values in-dicated by open circles nearly agree with the calculated ones.

As discussed in the previous section, the sidewall angle ofthe semiconductor is more crucial for the DBR reflectivitythan the sidewall roughness. We calculated the angle depen-dence of the reflectivity atλ = 1.55µm and at the wavelength

reflectivity of the etched facet and the DBR by fitting a theo-retical curve with the minimum threshold current density foreach type of device. We fabricated simple stripe lasers with-out DBRs but with a set of etched facets, along with DBRlasers. We first evaluated the reflectivity of the etched facetby measuringJth for these lasers. It was 28% and 10% for de-vices A and B, respectively. Based on these results, the DBRreflectivity of device A, which recorded the lowestJth, was

Fig. 5. FDTD calculation model.

100

0

20

60

80

40

1.3 1.4 1.5 1.6 1.7 1.8Wavelength [µm]

Ref

lect

ivity

RD

BR

[%]

Device A

Device B

Fig. 6. Reflection spectra of DBRs. Solid and dashed curves indicate the-oretical values for devices A and B, respectively. Open and closed circlesindicate experimental values for these devices.

100

9085 86 87 88 89

80

60

40

20

0

Sidewall Angle θ [deg.]

Ref

lect

ivity

RD

BR

[%]

RDBR @ λ = 1.55µm

PeakRDBR

θ

Fig. 7. DBR reflectivity calculated with sidewall angle.

−30

−25

−20

−15

−10

−5

0

Intensity[dB]

+0 fs

+40 fs

+80 fs

Fig. 8. Intensity profiles around DBR in device A.

Jpn. J. Appl. Phys. Vol. 39 (2000) Pt. 1, No. 6A M. ARIGA et al. 3409

as 3.2 mA/µm. The evaluated reflectivity was 85%, whichagreed well with the theoretical value obtained by the FDTDmethod. We experimentally and theoretically confirmed thatthe tilt of the vertical sidewall was more crucial for obtaininga high-reflectivity DBR than the roughness of the sidewall.

Acknowledgments

We would like to thank Professor Y. Kokubun, Professor Y.Hirose and Associate Professor H. Arai, Yokohama NationalUniversity, for their valuable discussions.

1) T. Baba, M. Hamasaki, N. Watanabe, P. Kaeplung, A. Matsutani, T.Mukaihara, F. Koyama and K. Iga: Jpn. J. Appl. Phys.35 (1996) 1390.

2) M. Hamasaki, P. Kaewplung, T. Baba, A. Matsutani, F. Koyama and K.Iga: Proc. Optelectronic and Communication Conf., Makuhari (1996)18P-31.

3) Y. Yuan, T. Brock, P. Bhattacharya, C. Caneau and R. Bhat: IEEE Pho-ton. Technol. Lett.9 (1997) 881.

4) T. Mukaihara, N. Yamanaka, N. Iwai, T. Ishikawa and A. Kasukawa:Electron. Lett.34 (1998) 882.

5) E. Höfling, R. Werner, F. Schäfer, J. P. Reithmaier and A. Forchel: Elec-tron. Lett.35 (1999) 154.

6) M. M. Raj, Y. Saka, J. Wiedmann, H. Yasumoto and S. Arai: Jpn. J.Appl. Phys.38 (1999) L1240.

7) A. Matsutani, H. Ohtsuki, F. Koyama and K. Iga: Jpn. J. Appl. Phys.38(1999) 4260.

8) R. Jambunathan and J. Singh: IEEE J. Quantum Electron.33 (1997)1180.

9) N. Habu, H. Tago, T. Baba, A. Matsutani, F. Koyama and K. Iga: Ext.Abstr. Fall Meet. Japan Society of Applied Physics, 1996, 8p-KH-16[in Japanese].

plane and roughness of less than 10 nm. In a GaInAsP/InPstripe laser system with such a DBR at one end, the minimumthreshold current normalized by the stripe width was as low

timated the amount of light power scattered in each directionby the DBR. 85% light power was reflected and coupled tothe waveguide, as mentioned above, and 9.2% was reflectedbut not coupled to the waveguide. This causes a decreasein the external quantum efficiency. In addition, 3.5% is de-flected upward, 2.0% is deflected downward, and 0.3% passesthrough the DBR to the right. Therefore, even if we install amonitor detector on the right hand side of such a DBR, it willstill be difficult to obtain high efficiency. We have reportedthat the diffraction loss can be significantly reduced by nar-rowing the air space toλ/4, i.e., 0.39µm for λ = 1.55µm.9)

However, this narrow space is difficult to fabricate without se-vere optimization of etching conditions. We also determinedthat, even though the air spaces are designed to be 3λ/4, thediffraction loss in the spaces can be suppressed and the max-imum reflectivity can be improved to>90% by plugging thespaces with some transparent material such as a polyimidewith a refractive index of∼1.6. High-performance lasinghas been demonstrated by plugging the spaces with benzo-cyclobutene (BCB) polymer.6)

5. Conclusions

We fabricated a high reflectivity semiconductor/air DBRby ICP etching with pure Cl2 gas. The DBR had smooth side-walls with an angle of more than 89◦ against the substrate